Battlefield V crashes at launch with dxr enabled

[Idcrafter]

it always crashes when dxr is enabled

[doitsujin]

Thank you very much for providing such detailed information about your hardware, driver, operating system, and vkd3d-proton logs. We can totally fix this based on the report alone!

[Idcrafter]

i have a rtx 2060 with 510 nvidia drivers ryzen 5 3600 with 48gb ddr4 ddr3600 cl16 Asus B450-f mainboard arch 5.16.11 KDE

[ProjectSynchro]

Thank you very much for providing such detailed information about your hardware, driver, operating system, and vkd3d-proton logs. We can totally fix this based on the report alone!

I just had a go at reproducing this (granted I don't have NVIDIA hardware), and I do indeed get a crash. I'm currently running: RX 6900XT Mesa 23.0.0-git commit:6cab999 Fedora 36 Workstation (Kernel 5.18.4-xm1.0 XanMod) VKD3D-proton 2.6.0 (Whatever is currently shipping with Proton Experimental Appears to be commit SHA b849bd425613afa)

I initially ran the game via Steam using the launch option RADV_PERFTEST=rt VKD3D_CONFIG=dxr %command% in order to expose the correct ray tracing features.

Enabling DX12 and DXR in the game settings causes the game to crash after restarting the game to apply changes. It craps out just as it starts setting the window size.

With VKD3D_DEBUG=err set I get the following in my log 0848:err:d3d12_state_object_parse_subobjects: Must have shader config.. I attached a log file with VKD3D_DEBUG set to warn as there is a bunch more noise in this log.

I do know that in Windows, this game does not work correctly with AMD hardware as the game may depend on some NVIDIA specific stuff for ray tracing. (See this post on the EA help forums)

Let me know if I can provide more information or better logs. (I don't know how helpful the one I provided is.)

[jkrhu]

The problem is still occuring. Crashes during the initial loading when DXR is enabled. DX11 and DX12 is fine, as long as DXR is off. Tried changing the config to just =dxr, removing the FEATURE_LEVEL, but makes no difference.

Also, on a different note while we're at it, the game likes to run out of swap file on DX12, when it caches shaders on every new map. It's actually impossible to not experience it when none of the maps were played before. The game is also so bad at this, literally every map and every piece of geometry with rare exceptions creates huge stutters. Usually after 1-2 matches, it fills up the 8GB swap file completely and crashes along with Steam itself. I'm sure it's related to the cache, because subsequent matches on the cached maps don't allocate that much swap file. Another thing to point out is that at that stage, I still have RAM left, but it prefers to use swap file either way. After compiling shaders fully, I still ran out of swap during one round on a map I haven't cached before. I believe the game doesn't purge shader caches it doesn't need anymore and piles them on causing OOM on my system, sometimes after one or two rounds.

Specs: Fedora 36, 6.0.5 kernel RTX 3060 Driver 520.56.06 i5 10400 16GB RAM

Commands: VKD3D_CONFIG=dxr11 VKD3D_FEATURE_LEVEL=12_1 PROTON_ENABLE_NVAPI=1 PROTON_HIDE_NVIDIA_GPU=0

Attaching the log file below: BFV DXR Nvidia.log

[K0bin]

0af8:err:d3d12_state_object_pipeline_data_find_global_state_object: Mismatch in inherited associations for kind 1.

This is probably the interesting one.

            /* To hypothetically support this for global root signatures,
             * we'd have to partition any RTPSO into N VkPipelines and select
             * the appropriate pipeline at DispatchRays() time based on the currently bound root signature ...
             * Leave this as a TODO until we observe that applications rely on this esoteric behavior. */

That seems to be the case for BF5.

[Nlk29]

That seems to be the case for BF5.

is there anything i can do about that? Fucking h8 windows and don't wanna go back to it.

[HansKristian-Work]

I would have to somehow implement that API wart. It's really hard and I don't know when I'll be able to, but I'll have to do it eventually.